Measurement of light-by-light scattering inATLAS

Marcin Guzik

AGH University of Science and Technology, Cracow

Seminarium środowiskowe Fizyki Cząstek Białasówka21 April 2017

1/29

Electromagnetic interactions in Pb+Pb collisions

2RPb < b

[Fermi, Nuovo Cim. 2 (1925) 143][Weizsacker, Z. Phys. 88 (1934) 612][Williams, Phys. Rev. 45 (10 1934) 729]

Equivalent Photon Approximation (EPA)

σEPAA1A2(γγ)→A1A2X =

∫∫dω1dω2 n1 (ω1)n2 (ω2)σγγ→X (Wγγ)

with n (b, ω) = Z2αemπω

∣∣∣∣∫ dq⊥q2⊥F(Q2)Q2 J1 (bq⊥)

∣∣∣∣2Q2 < 1

R2 and ωmax ≈ γR

2/29

LHC as a photon-photon colliderpp collisions Pb+Pb collisions

Pros

• harder EPA γ spectrum(ωmax ∼ TeV)

• more data available(∼ 35 fb−1

)Cons

• large pile-up (multipleinteractions per bunch crossing)

• problems with triggering on lowpT objects

Pros

• AA (γγ) x-sec ∝ Z4

• gluonic x-sec ∝ A2

⇒ lower QCD bkg.

• low pile-up (< 1%)

Cons

• softer EPA γ spectrum(ωmax ∼ 0.1TeV)

• relatively small datasample

[PR D94 (2016) 3, 032011]

[ALICE Collaboration,EPJC 73 (2013) 2617]

3/29

Search for light-by-light scattering[arXiv:1702.01625]

Motivation

• first direct observation ofγγ → γγ scattering

• previous indirect measurements used:a) multi-photon Breit-Wheeler

reaction(ω + nω0 → e+e−) [PRL 79 (1997) 1626]

b) photon splittingc) Delbruck scattering

Pb Pb

γ γ

γγ

Pb Pb

4/29

Positron Production in Multiphoton Light-by-Light Scattering[Physical Review Letters 79, 1626 (1997)]

95% CL on the residualbackground from showersof lost beam particles af-ter subtracting the laser-offpositron rate

simulation of trident processω + nω0 → e′e+e−

power law fittedto the data

Abstract

5/29

Scattering of 1.3 Mev Gamma-Rays by an Electric Field[Phys. Rev. 90, 720 (1953)]

6/29

Experimental investigation of high-energy photon splitting in atomic fields[Phys. Rev. Lett. 89 (2002) 061802], [arXiv:hep-ex/0111084]

• source: Compton scatteredlaser light

• Compton scattered electron Emeasured

• both final γ detected

7/29

Standard Model Predictions

Recent SM Predictions for the LHC[A. Szczurek et al. PRC 93 (2016) 4, 044907], [D. d’Enterria et al. PRL 111 (2013) 080405]

8/29

The LHC and the ATLAS detector

2015 Pb+Pb dataL ≈ 0.5 nb−1

√sNN = 5.02 TeV

9/29

The ATLAS detector components

ηTracking in 2T Solenoid

MuonsPhotons

2015 Pb+Pb dataL ≈ 0.5 nb−1

√sNN = 5.02 TeV

10/29

LbyL - Photon Identification

γ cuts: ET > 3 GeV, |η| < 2.37

Shower shape variables used to γ PID

• Eratio ≡ ratio of the energy difference associated with thelargest and second largest energy deposits to the sum of thesedeposits in the first layer of EM calo

• f1 ≡ fraction of energy reconstructed in the first layer withrespect to the total energy of the cluster

• Weta2 ≡ lateral width of the shower in the middle layer

γπ0

11/29

Search for light-by-light scattering[arXiv:1702.01625]

Trigger

• total ET in calorimeterbetween 5 and 200 GeV

• no more than one hit ininner MBTS

• less than 10 hits in the pixeldetector

Main sources of bkg.

• Central ExclusiveProduction (CEP) gg→ γγ

• misidentification of electronsfrom γγ → ee

Event Selection

• two photons withET > 3 GeV, |η| < 2.37

• no tracks from IP

• mγγ > 6 GeV, pγγT < 2 GeV

• Aco =(

1− ∆φπ

)< 0.01

acoplanarity cutγγ0 0.01 0.02 0.03 0.04 0.05 0.06

γγC

EP

+

Nsi

gN

/ si

gN

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

ATLAS

= 5.02 TeVNNsPb+Pb

< 2 GeVγγT

p

background MCγγsignal + CEP

Pb Pb

γ

γ

Pb Pb

12/29

More about background

acoplanarityγγ0.02 0.04 0.06 0.08 0.1 0.12

Events/0.01

0

1

2

3

4

5

6

7

8

9

10ATLAS

= 5.02 TeVNNsPb+Pb⩾1

ZDCData, n

MCγγCEP

CEP of γγ

• normalisation from fitto the data in regionof Aco > 0.02

• ±0.01 as a systematic

• ZDC based studies asa cross check

13/29

More about background γγ → e+e−

[GeV]-e+em0 5 10 15 20 25 30 35 40 45 50

Eve

nts

/ GeV

0

100

200

300

400

500

600

700

800ATLAS

=5.02 TeVNNsPb+Pb

| < 2.47eη > 2.5 GeV, |eTE

-1bµData, 480

MC -e+e→γγ

[GeV]-e+e

Tp

0 1 2 3 4 5

Eve

nts

/ 0.2

GeV

0

100

200

300

400

500

600

700

800

900ATLAS

=5.02 TeVNNsPb+Pb

| < 2.47eη > 2.5 GeV, |eTE

-1bµData, 480

MC -e+e→γγ

eη3− 2− 1− 0 1 2 3

Ele

ctro

ns /

0.1

0

100

200

300

400

500

600ATLAS

=5.02 TeVNNsPb+Pb

| < 2.47eη > 2.5 GeV, |eTE

-1bµData, 480

MC -e+e→γγ

[GeV]eTE

0 5 10 15 20 25

Ele

ctro

ns /

0.5

GeV

0

200

400

600

800

1000

1200 ATLAS

=5.02 TeVNNsPb+Pb

| < 2.47eη > 2.5 GeV, |eTE

-1bµData, 480

MC -e+e→γγ

14/29

More about background γγ → e+e−

acoplanarityγγ0 0.1 0.2 0.3 0.4 0.5 0.6

Eve

nts

/ 0.0

2

0

5

10

15

20

25

30ATLAS

=5.02 TeVNNsPb+Pb

= 1 control regiontrkN

-1bµData, 480 MC-e+e→γγ

10× MC γγ→γγ

acoplanarityγγ0 0.1 0.2 0.3 0.4 0.5 0.6

Eve

nts

/ 0.0

2

0

5

10

15

20

25

30ATLAS

=5.02 TeVNNsPb+Pb

= 2 control regiontrkN

-1bµData, 480 MC-e+e→γγ

10× MC γγ→γγ

• precision limited by the data statistic

• conservative uncertainty of 25% assigned to MC

15/29

Search for light-by-light scattering[arXiv:1702.01625]

16/29

Trigger and photon PID efficiency studies

Trigger efficiency (γγ → e+e−)

• independent trigger: coincidence of signals in both ZDC sides anda requirement on the total ET in the calorimeter below 50 GeV.

• events with only two reconstructed tracks and two EM energyclusters (with cl Aco < 0.2)

• about 70% at(Ecl1T + Ecl2T

)= 6 GeV to 100% above 9 GeV

• error function parametrisation used to reweight the MC• MBTS veto studied using supporting trigger (98± 2)%

γ PID (γγ → e+e−)

• one photonand two tracks

• FSR selection:pttγT < 1 GeV

[GeV]TE

0 2 4 6 8 10 12 14 16 18

Pho

ton

PID

effi

cien

cy

0

0.2

0.4

0.6

0.8

1

1.2

1.4ATLAS

=5.02 TeVNNsPb+Pb

-1bµData, 480

MC

17/29

Photon reconstruction efficiency studies

[GeV]γTE

0 2 4 6 8 10 12 14 16 18 20

Eve

nts

/ GeV

0

20

40

60

80

100

120ATLAS

=5.02 TeVNNsPb+Pb

=1γ=1, NeN

< 2 GeVtrk2

T> 5 GeV, pe

TE

-1bµData, 480

MC -e+e→γγ

[GeV]γ-e+e

Tp

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Eve

nts

/ 0.2

GeV

0

20

40

60

80

100

120ATLAS

=5.02 TeVNNsPb+Pb

=1γ=1, NeN

< 2 GeVtrk2

T> 5 GeV, pe

TE

-1bµData, 480

MC -e+e→γγ

γ reco (γγ → e+e−) withhard bremsstrahlung

• one electron withET > 5 GeV andtwo good qualitytracks

• unmatched track:ptrk2T < 2 GeV

[GeV]trk2T

- peTE

0 2 4 6 8 10 12 14 16 18

Pho

ton

reco

nstr

uctio

n ef

ficie

ncy

0

0.2

0.4

0.6

0.8

1

1.2

1.4ATLAS

=5.02 TeVNNsPb+Pb

-1bµData, 480

MC

18/29

Search for light-by-light scattering[arXiv:1702.01625]

Photon Performance Studies(done with γγ → l+l− events)

• trigger efficiency studies

• γ reconstruction with hardbremsstrahlung

• γ PID with FSR radiation

• γ energy scale andresolution

Systematic Uncertainty

dominated by:

• γ reco

• γ PID

[GeV]cluster2T + Ecluster1

TE0 5 10 15 20 25 30 35 40 45 50

Eve

nts

/ 0.5

GeV

0

50

100

150

200

250

300ATLAS

=5.02 TeVNNsPb+Pb

| < 2.47eη > 2.5 GeV, |eTE

-1bµData, 480

/ndf=150/89)2χdefault MC (

/ndf=210/89)2χ scale +5% (TE

/ndf=285/89)2χ scale -5% (TE

Source of uncertainty Relative uncertainty

Trigger 5%Photon reco efficiency 12%Photon PID efficiency 16%Photon energy scale 7%Photon energy resolution 11%

Total 24%19/29

Search for light-by-light scattering[arXiv:1702.01625]

Results: data - 13 event, expected - 7.3 signal and 2.6 bkg. eventsSelection γγ → e+e− CEP gg → γγ Hadronic Other Total Signal Data

fakes fakes background

Preselection 74 4.7 6 19 104 9.1 105Ntrk = 0 4.0 4.5 6 19 33 8.7 39pγγT < 2 GeV 3.5 4.4 3 1.3 12.2 8.5 21Aco < 0.01 1.3 0.9 0.3 0.1 2.6 7.3 13

Uncertainty 0.3 0.5 0.3 0.1 0.7 1.5

acoplanarityγγ0 0.01 0.02 0.03 0.04 0.05 0.06

Eve

nts

/ 0.0

05

0

2

4

6

8

10

12

14 ATLAS

= 5.02 TeVNNsPb+Pb

Signal selectionno Aco requirement

-1bµData, 480 MC γγ→γγ

MC -e+e→γγ MC γγCEP

|γγ

|y0 0.5 1 1.5 2 2.5 3

Eve

nts

/ 0.5

0

2

4

6

8

10

12

14 ATLAS

=5.02 TeVNNsPb+Pb

Signal selectionwith Aco < 0.01

-1bµData, 480 MC γγ→γγ

MC -e+e→γγ MC γγCEP

20/29

Search for light-by-light scattering[arXiv:1702.01625]

Results:• significance of 4.4σ estimated using profile likelihood

method (expected significance of 3.8σ)• x-sec measured in fiducial region of pγT > 3 GeV, |ηγ | < 2.4,

mγγ > 6 GeV, pγγT < 2 GeV, Aco < 0.01σ = 70± 20 (stat.)± 17 (syst.)nbSM predictions: 45± 9 nb ([PRL 111 (2013) 080405]), 49± 10 nb ([PRC 93 (2016) no.4, 044907])

γγ→γγµ

0 0.5 1 1.5 2 2.5 3

ln L

∆-

2

0

5

10

15

20

25

Observed

SM expected

ATLAS

=5.02 TeVNNsPb+Pb

-1bµL dt = 480 ∫

σ1

σ2

σ3

σ4

σ5

[GeV]γγT

p0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Eve

nts

/ GeV

0

2

4

6

8

10

12

14

16

18

20ATLAS

=5.02 TeVNNsPb+Pb

Signal selectionwith Aco < 0.01

-1bµData, 480 MC γγ→γγ

MC -e+e→γγ MC γγCEP

21/29

BSM Physics in HI - Axion Search With UPC[arXiv:1607.06083]

La = 12 (∂a)2 − 12m

2aa2 − 14

aΛFF

Pb Pb

γ γaγγ

Pb Pb

expected axion searches sensitivity

22/29

BSM Physics in HI - Axion Search With UPC[arXiv:1607.06083]

La = 12 (∂a)2 − 12m

2aa2 − 1

4 cos2 θWaΛBB

Pb Pb

γ γaγγ

Pb Pb

expected axion searches sensitivity

23/29

BSM Physics in HI - Axion Search With UPC[arXiv:1607.06083]

La = 12 (∂a)2 − 12m

2aa2 − 14

aΛFF

Pb Pb

γ γaγγ

Pb Pb

σALP = 5 nb, ma = 15 GeV and 40 GeVassumed E resolution of 0.5 GeV

24/29

BSM Physics in HI - Axion Search With UPC

[GeV]γγm0 5 10 15 20 25 30

Eve

nts

/ 0.5

GeV

0

2

4

6

8

10

12 ATLAS

=5.02 TeVNNsPb+Pb

Signal selectionAco < 0.01

-1bµData, 480 MC γγ→γγ

MC -e+e→γγ MC γγCEP

25/29

LbyL Scattering Constraint on Born-Infeld Theory[arXiv:1703.08450]

LQED = −14FµνFµν → LBI = β2

(1−

√1 + 1

2β2FµνFµν − 16β4

(FµνFµν

)2)

26/29

LbyL Scattering Constraint on Born-Infeld Theory[arXiv:1703.08450]

• limit on Born-Infeld scale M =√β & 100 GeV 5 orders of

magnitude grater than the previous one from PVLAS• in case of Born-Infeld SM extention with U (1)Y realized

nonlineary MY = cos θWM & 90 GeV• such theory has finite-energy electroweak monopole solution

less constrained by higgs than in other extensions of SMwhich because of MY → Mmonopole & 11 TeV is out of reachat the LHC

27/29

Summary

• The first direct evidence for γγ → γγ scattering withsignificance of 4.4σ has been reported.• improvements in the precision expected with more Pb+Pb

data to be collected in 2018

• This result has already been used by J. Ellis etal. [arXiv:1703.08450] to derive the limits on Born-Infeld theory

• According to S. Knapen et al. [arXiv:1607.06083] UPC data can beused in BSM searches eg. for ALP

28/29

Thank You for Your Attention!

29/29